CN110781021B

CN110781021B - Anomaly detection method and device, computer equipment and storage medium

Info

Publication number: CN110781021B
Application number: CN201911039595.0A
Authority: CN
Inventors: 赵菁; 荆彦青; 张力柯; 王君乐; 王逸超; 张远斌
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2019-10-29
Filing date: 2019-10-29
Publication date: 2021-07-06
Anticipated expiration: 2039-10-29
Also published as: CN110781021A

Abstract

The application discloses an anomaly detection method and device, computer equipment and a storage medium, and belongs to the technical field of computers. According to the application, a plurality of test values of the target prop parameter of the interactive prop in different use processes are obtained, the use state information and the stability information of the target prop parameter are quantized according to the test values, so that whether the target prop parameter is abnormal or not is detected according to the use state information and the stability information, the influence of user subjective factors in an abnormal detection process can be avoided, the target prop parameter causing abnormity is accurately detected, and the application program providing the interactive prop is optimized.

Description

Anomaly detection method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to an anomaly detection method and apparatus, a computer device, and a storage medium.

Background

With the development of computer technology and the diversification of terminal functions, more and more games can be played on the terminal. Among them, the First-Person Shooting game (FPS) is a popular game, and various firearm interactive props can be provided in the FPS game.

The property parameters of the firearm-type interactive property determine the 'firearm feel' of the user when the user uses the property, for example, the property parameters include recoil, firing speed, firing interval, range and the like, and firearm-type interactive properties with different property parameters may bring different firearm feels to the user.

At present, due to the reason of game data confidentiality, the measurement of the handfeel of firearms according to the subjective feelings of users and game feedback is a main analysis method of the handfeel of firearms, so that whether the handfeel of firearms is stable can be only reflected qualitatively, different terminals can influence the judgment of users on the handfeel of firearms due to the obvious difference of the subjective feelings of the users, and once the handfeel of firearms is abnormal, the abnormal reasons are difficult to find through the game feedback of the users. Therefore, it is desirable to provide an abnormal detection method for analyzing the hand feeling of a firearm by a user, so as to accurately test the cause of the abnormal hand feeling of the firearm, thereby optimizing a shooting game.

Disclosure of Invention

The embodiment of the application provides an anomaly detection method and device, computer equipment and a storage medium, and can solve the problems of low accuracy of firearm hand feeling analysis and difficulty in anomaly detection. The technical scheme is as follows:

in one aspect, an anomaly detection method is provided, and the method includes:

obtaining a plurality of test values of target prop parameters of the interactive props in different using processes, wherein the target prop parameters are used for reflecting target interactive characteristics of the interactive props in the using processes;

determining the use state information and the stability information of the target prop parameter according to each test value of the target prop parameter;

and detecting whether the target prop parameter of the interactive prop is abnormal or not according to the use state information and the stability information.

In a possible embodiment, the obtaining a plurality of test values of the target item parameter of the interactive item in different use processes includes:

acquiring a plurality of data sets generated by the interactive prop in different use processes, wherein each data set comprises a plurality of data items generated when the interactive prop with the same prop parameter is used for multiple times;

and determining the plurality of test values of the target prop parameter according to each data item of each data group.

In one possible embodiment, the data item is a target image, and the acquiring a plurality of data sets generated by the interactive prop in different use processes includes:

the method comprises the steps that an interactive prop with different prop parameters is tested for multiple times through a configuration file to obtain a plurality of image sets, and each image set comprises a plurality of test images generated when the interactive prop with the same prop parameters is used for multiple times;

extracting a target area from each test image of the plurality of image sets, and acquiring a binary image of each target area;

and carrying out abnormality detection on each binary image, deleting the abnormal images in each binary image to obtain each target image, and determining the target image corresponding to the same image set as one data set in the plurality of data sets.

In one possible embodiment, the plurality of image sets are named with respective timestamps and prop parameters when they are stored.

In one aspect, an abnormality detection apparatus is provided, the apparatus including:

the acquisition module is used for acquiring a plurality of test values of a target prop parameter of the interactive prop in different use processes, wherein the target prop parameter is used for reflecting a target interactive characteristic of the interactive prop in the use process;

the determining module is used for determining the use state information and the stability information of the target prop parameter according to each test value of the target prop parameter;

and the detection module is used for detecting whether the target prop parameter of the interactive prop is abnormal or not according to the use state information and the stability information.

In one possible implementation, the obtaining module includes:

the first acquisition unit is used for acquiring a plurality of data sets generated by the interactive prop in different use processes, wherein each data set comprises a plurality of data items generated when the interactive prop with the same prop parameter is used for multiple times;

and the determining unit is used for determining the plurality of test values of the target prop parameter according to each data item of each data group.

In one possible implementation, the data item is a target image, and the first acquiring unit includes:

the system comprises a testing subunit, a processing unit and a processing unit, wherein the testing subunit is used for testing interactive props with different prop parameters for multiple times through configuration files to obtain a plurality of image sets, and each image set comprises a plurality of testing images generated when the interactive props with the same prop parameters are used for multiple times;

the obtaining subunit is configured to extract a target region from each test image of the plurality of image sets, and obtain a binary image of each target region;

and the first determining subunit is used for carrying out abnormality detection on each binary image, deleting the abnormal images in each binary image to obtain each target image, and determining the target image corresponding to the same image set as one data set in the plurality of data sets.

In one possible embodiment, the obtaining subunit is configured to:

acquiring a target region and a filtering region corresponding to the plurality of image sets, wherein the filtering region is located in the target region;

performing target area cutting and gray level processing on each test image of the image sets to obtain a gray level image of a target area in each test image;

and carrying out binarization processing on the gray level image of each target area, and setting pixel points covered by the filtering area to be 0 in each binarized image to obtain a binary image of each target area.

In one possible embodiment, the apparatus is further configured to:

randomly selecting at least one binary image from each image set to test, if the test result does not meet the target condition, adjusting the processing parameters of the binarization processing process, and repeatedly executing the steps of obtaining the binary images and carrying out abnormity detection until the test result meets the target condition, wherein the target condition is used for representing the acceptable range of errors generated by the processing parameters.

In one possible embodiment, the determining unit is configured to:

if the target prop parameter is the maximum offset between the shooting position and the target position of the interactive prop, selecting a reference point in any target image;

determining at least one target pixel point which is farthest from the reference point in each direction from pixel points with pixel values of 1 in the target image;

and determining the distance between the at least one target pixel point and the reference point in each direction as the maximum offset in each direction, and determining the maximum offset in each direction corresponding to the target image as a test value of the target prop parameter.

In one possible implementation, the reference point is a geometric center of a test image corresponding to the target image; or, the reference point is a pixel point of which the lowest pixel value in the target image is 1.

In one possible embodiment, the test subunit is configured to:

in an application program, selecting an interactive prop with the prop parameters configured by the configuration file;

determining a fixed shooting position and a shooting target for a virtual object in a virtual scene provided by the application program;

controlling the virtual object to continuously shoot at the shooting position to the shooting target, and carrying out image interception on the virtual scene within the target duration to obtain a test image;

and repeatedly executing the steps of controlling the virtual object to shoot continuously and obtaining the test images to obtain a plurality of test images with the same prop parameters, and determining the test images as an image set.

In one possible embodiment, the determining module comprises:

the second obtaining unit is used for obtaining at least one of an average value or a standard deviation of each test value, and determining at least one of the average value or the standard deviation as the use state information of the target prop parameter;

and the analysis unit is used for acquiring a plurality of comparison groups corresponding to the plurality of data groups where the test values are located, and performing statistical analysis on the plurality of data groups and the plurality of comparison groups to obtain the stability information of the target prop parameter.

In a possible embodiment, the analysis unit comprises:

the displacement inspection subunit is used for performing displacement inspection on the plurality of data groups and the plurality of comparison groups to obtain a distribution result of the target prop parameter;

the self-service method statistics subunit is used for performing self-service method statistics on the plurality of data groups and the plurality of comparison groups to obtain the difference degree of the target prop parameters;

a second determining subunit, configured to determine at least one of the distribution result or the degree of difference as the stability information.

In one possible embodiment, the permutation check subunit is configured to:

for any data group and a comparison group of the data groups, acquiring a target mean value difference of test values of target prop parameters between the data group and the comparison group;

randomly dividing each test value of the data group and the comparison group into two different permutation groups, and determining the average value difference between the two permutation groups;

repeating the steps of randomly dividing the permutation group and determining the average value difference for multiple times to obtain multiple average value differences, and acquiring a first sampling distribution formed by the multiple average value differences;

and acquiring a probability value of the permutation test according to a confidence interval of the target mean value difference falling in the first sampling distribution, and determining a distribution result of the target prop parameter based on the probability value.

In one possible embodiment, the self-service statistics subunit is configured to:

for any data group and a comparison group corresponding to the data group, obtaining the probability value of the data group and the comparison group in replacement test;

randomly extracting test values from the comparison group to form a test group, and performing replacement test on the data group and the test group to obtain a test probability value;

repeatedly executing the steps of forming a test group and obtaining test probability values to obtain a plurality of test probability values, and obtaining second sampling distribution formed by the test probability values;

and acquiring the difference degree of the target prop parameter according to the occurrence frequency of the probability value in the designated interval in the second sampling distribution.

In one aspect, a computer device is provided that includes one or more processors and one or more memories having at least one program code stored therein, the at least one program code being loaded by the one or more processors and executed to implement the operations performed by the anomaly detection method according to any one of the possible implementations described above.

In one aspect, a storage medium is provided, in which at least one program code is stored, the at least one program code being loaded and executed by a processor to perform operations performed to implement the anomaly detection method according to any one of the above possible implementations.

The beneficial effects brought by the technical scheme provided by the embodiment of the application at least comprise:

the target prop parameters of the interactive props are obtained through a plurality of test values in different using processes, using state information and stability information of the target prop parameters are quantized according to the test values, whether the target prop parameters are abnormal or not is detected according to the using state information and the stability information, the influence of user subjective factors in an abnormal detecting process can be avoided, the target prop parameters causing the abnormality are accurately detected, and therefore the application program providing the interactive props is optimized.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic diagram of an implementation environment of an anomaly detection method according to an embodiment of the present application;

fig. 2 is a flowchart of an anomaly detection method provided in an embodiment of the present application;

FIG. 3 is a flow chart of acquiring a data set according to an embodiment of the present application;

FIG. 4 is a schematic interface diagram of a test image provided in an embodiment of the present application;

FIG. 5 is a schematic diagram of a test image naming method provided by an embodiment of the present application;

FIG. 6 is a schematic diagram of a target area and a filter area provided by an embodiment of the present application;

FIG. 7 is a schematic diagram of a target area and a filter area provided by an embodiment of the present application;

fig. 8 is a schematic diagram of an extraction diagram of a bullet hole provided in an embodiment of the present application;

FIG. 9 is a schematic diagram of an anomaly image provided by an embodiment of the present application;

FIG. 10 is a flow chart for analyzing maximum offset according to an embodiment of the present disclosure;

FIG. 11 is a schematic illustration of a selected datum provided by an embodiment of the present application;

FIG. 12 is a schematic view of a selected datum provided by an embodiment of the present application;

FIG. 13 is a schematic diagram of an upward maximum offset provided by an embodiment of the present application;

FIG. 14 is a schematic illustration of a first sampling profile provided by an embodiment of the present application;

FIG. 15 is a schematic diagram of a second sampling profile provided by an embodiment of the present application;

FIG. 16 is an enlarged view of a portion of a second sampling profile provided by an embodiment of the present application;

FIG. 17 is a schematic illustration of a second sampling profile provided by an embodiment of the present application;

FIG. 18 is an enlarged partial view of a second sampling profile provided by an embodiment of the present application;

FIG. 19 is a schematic diagram of a method for analyzing a feel of a firearm according to an embodiment of the present disclosure;

FIG. 20 is a schematic diagram of a firearm feel analysis system provided by an embodiment of the present application;

FIG. 21 is a schematic diagram of a firearm feel analysis system provided by an embodiment of the present application;

FIG. 22 is a schematic diagram of a method for analyzing the feel of a firearm according to an embodiment of the present disclosure;

fig. 23 is a schematic structural diagram of an abnormality detection apparatus according to an embodiment of the present application;

fig. 24 is a schematic structural diagram of a computer device according to an embodiment of the present application.

Detailed Description

To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

Hereinafter, terms related to the present application are explained.

Virtual scene: is a virtual scene that is displayed (or provided) by an application program when the application program runs on a terminal. The virtual scene may be a simulation environment of a real world, a semi-simulation semi-fictional virtual environment, or a pure fictional virtual environment. The virtual scene may be any one of a two-dimensional virtual scene, a 2.5-dimensional virtual scene, or a three-dimensional virtual scene, and the dimension of the virtual scene is not limited in the embodiment of the present application. For example, a virtual scene may include sky, land, ocean, etc., the land may include environmental elements such as deserts, cities, etc., and a user may control a virtual object to move in the virtual scene.

Virtual object: refers to a movable object in a virtual scene. The movable object can be a virtual character, a virtual animal, an animation character, etc., such as: characters, animals, plants, oil drums, walls, stones, etc. displayed in the virtual scene. The virtual object may be an avatar in the virtual scene that is virtual to represent the user. The virtual scene may include a plurality of virtual objects, each virtual object having its own shape and volume in the virtual scene and occupying a portion of the space in the virtual scene.

Alternatively, the virtual object may be a Player Character controlled by an operation on the client, an Artificial Intelligence (AI) set in the virtual scene fight by training, or a Non-Player Character (NPC) set in the virtual scene interaction. Alternatively, the virtual object may be a virtual character playing a game in a virtual scene. Optionally, the number of virtual objects participating in the interaction in the virtual scene may be preset, or may be dynamically determined according to the number of clients participating in the interaction.

And (3) interaction of props: the interactive property can be a firearm interactive property such as a machine gun, a pistol and a rifle, and the type of the interactive property is not specifically limited in the application.

Optionally, the interactive prop may have different prop parameters, each of which is used to reflect the interactive characteristics of the interactive prop during use, for example, the prop parameters of the firearm-type interactive prop may include: at least one of firearm accessories, recoil, firing rate, firing interval, reloading speed, injury, range, portability, or payload.

The firearm accessories refer to matching tools of firearm interactive props, such as a telescope, a muzzle, a handle, a stock and the like.

The recoil refers to the negative influence of the accuracy of the firearm interaction prop after shooting or continuous shooting, and the larger the recoil is, the higher the amplitude of the accuracy reduction caused after the shooting or continuous shooting is.

Wherein, shooting speed refers to the speed of shooting bullets in unit time in the process of continuously shooting by the firearm interactive props.

The firing interval refers to the minimum time interval between the two adjacent bullets launched by the firearm type interactive prop in the continuous shooting process.

Wherein, the bullet changing speed refers to the time required by the firearm interactive prop when changing bullets.

The damage refers to the degree of damage that the firearm-like interactive prop can cause, for example, the degree of damage is expressed by the amount of damage caused.

The range refers to the farthest attack distance of the firearm interactive prop, and the damage of the firearm interactive prop in the range is attenuated along with the increase of the distance until the minimum damage is reached.

The portability of the firearm interactive prop is better, the more flexible the user controls the operation of the virtual object, and the faster the action speed of the virtual object.

Wherein, the loading capacity refers to the maximum bullet loading capacity of a cartridge clip of the firearm interaction prop.

Fig. 1 is a schematic diagram of an implementation environment of an anomaly detection method provided in an embodiment of the present application, and referring to fig. 1, the implementation environment includes: a test terminal 101 and a server 102.

The test terminal 101 is used for testing the interactive prop, and the test terminal 101 is provided with a configuration file and an application program. The application program can support providing of interactive props in a virtual scene, and the application program can be any one of a First-Person Shooting game (FPS), a third-Person Shooting game, a Multiplayer Online Battle sports game (MOBA), a virtual reality application program, a three-dimensional map program, a military simulation program or a Multiplayer gun type survival game.

The device types of the test terminal 101 may include: at least one of a smart phone, a tablet computer, an e-book reader, an MP3(Moving Picture Experts Group Audio Layer III, motion Picture Experts compression standard Audio Layer 3) player, an MP4(Moving Picture Experts Group Audio Layer IV, motion Picture Experts compression standard Audio Layer 4) player, a laptop portable computer, and a desktop computer. For example, the test terminal 101 may be a smart phone, or other handheld portable gaming device. The following embodiments are illustrated with the terminal comprising a smartphone.

Those skilled in the art will appreciate that the number of test terminals 101 described above may be greater or fewer. For example, the number of the test terminals 101 may be only one, or the number of the test terminals 101 may be several tens or hundreds, or more. The number and the device type of the test terminals 101 are not limited in the embodiment of the present application.

The server 102 may include at least one of a server, a plurality of servers, a cloud computing platform, or a virtualization center. Server 102 is configured to provide a background service for anomaly detection of the interactive prop. Alternatively, the server 102 may undertake primary computational work and the test terminal 101 may undertake secondary computational work; or, the server 102 undertakes the secondary computing work, and the test terminal 101 undertakes the primary computing work; alternatively, the server 102 and the test terminal 101 perform cooperative computing by using a distributed computing architecture.

The test terminal 101 and the server 102 are connected via a wireless network or a wired network.

In an exemplary scenario, a user logs in the test terminal 101, configures multiple sets of property parameters for one test run through a configuration file, and tests interactive properties with different property parameters on the test terminal 101 for multiple times to obtain test values of each property parameter in different use processes. The test terminal 101 may send the test value of each item parameter to the server 102, and the server 102 performs data processing and anomaly detection, and returns an anomaly detection result of each item parameter to the test terminal 101.

It should be noted that the test terminal 101 and the server 102 may be deployed on the same physical device, or may be deployed on different physical devices, and this embodiment of the present application does not specifically limit whether the two are located on the same physical entity.

Fig. 2 is a flowchart of an anomaly detection method according to an embodiment of the present application. Referring to fig. 2, taking the application of the embodiment to the server 102 as an example, the embodiment includes:

201. the server obtains a plurality of data sets generated by the interactive prop in different using processes, wherein each data set comprises a plurality of data items generated when the interactive prop with the same prop parameter is used for multiple times.

The data items may be in text form or image form. In the next embodiment, the data item is taken as an example of a target image, and a process of acquiring the data set is described in detail, which is not described herein again.

In the process, the server can collect data items of one or more user terminals in different using processes from the background, and the interactive props with the same prop parameters are divided into the same data group, so that the process of acquiring the data group can be simplified.

In some embodiments, the server may also directly perform multiple tests on the interactive prop to obtain the multiple data sets, so that different prop parameters can be configured individually, and a comparison set can be generated for each data set by controlling variables, thereby facilitating a subsequent anomaly detection process.

202. And the server determines a plurality of test values of the parameters of the target prop according to each data item of each data group.

The target property parameter is used for reflecting a target interaction characteristic of the interactive property in a use process, and the target property parameter may be a variable designated to be controlled in the current round of abnormality detection of the server, for example, the target property parameters of the data sets are different from each other, but the other property parameters except the target property parameter all have the same value.

Alternatively, the server may directly determine the respective data items as the respective test values. For example, when the parameter of the target prop is the payload, the data item may be a payload in a text form, and the server directly determines a data item as a test value.

In some embodiments, the server needs to perform data processing on each data item to obtain each test value, and the test values of different target property parameters may have different expressions, for example, each test value may be one value or a test value sequence formed by a plurality of values.

For example, when the target prop parameter is recoil, the recoil can be represented by using the maximum offset of the interactive prop, and the larger the maximum offset is, the stronger the recoil is, in this case, if the data item is a target image in an image form, the server can perform image processing on the target image, because the maximum offset is relative to the direction, the server can obtain the maximum offset of the interactive prop in different directions under each target image, and the server can determine the maximum offset (for example, the maximum offsets in four directions, namely, up, down, left, and right) of the same target image in different directions as a test value sequence of the recoil (that is, the target prop parameter).

In the step 201-202, the server obtains a plurality of test values of the target property parameter of the interactive property in different use processes, so that whether the target property parameter is abnormal or not can be detected through subsequent data analysis and processing steps on the basis of each test value.

203. And the server determines the use state information and the stability information of the target prop parameter according to each test value of the target prop parameter.

The use state information is used for reflecting the average use condition of the target prop parameter in multiple rounds of use processes, for example, the use state information may include an average value and a standard deviation of each test value, the average value is used for measuring the average size of the test values, the average size of the test values is larger when the average value is larger, the standard deviation is used for measuring the dispersion degree of each test value, and the randomness of the target prop parameter is larger when the standard deviation is larger.

The stability information is used to reflect whether the test values of the target prop parameter tend to be stable in multiple use processes, for example, the stability information may include a distribution result and a difference degree, the distribution result is used to indicate which sampling distribution each test value obeys, and the difference degree is used to indicate whether there is a significant difference between each test value.

In some embodiments, the server may obtain at least one of an average value or a standard deviation of each test value, and determine the at least one of the average value or the standard deviation as the usage status information of the target prop parameter; and acquiring a plurality of comparison groups corresponding to the plurality of data groups where the test values are located, and performing statistical analysis on the plurality of data groups and the plurality of comparison groups to obtain stability information of the target prop parameter. Therefore, the use state information of the target prop parameter can be analyzed, and the stability information of the target prop parameter can also be analyzed.

In step 202, since the server determines each test value according to each data item, each test value may also have a group corresponding to each data item, that is, each test value corresponds to a different data group, and therefore, the server may calculate an average value and a standard deviation of the test values in each data group by taking the data group as a unit. Furthermore, any two different data sets can be mutually compared, and when any data set is processed, one comparison set is selected for the data set, so that the data set and the corresponding comparison set can be subjected to statistical analysis, and the stability information of the data set can be obtained.

In some embodiments, the server may only obtain the use state information of the target prop parameter, or only obtain the stability information of the target prop parameter, so that the flow of data processing on each test value can be simplified.

204. And the server detects whether the target prop parameter of the interactive prop is abnormal or not according to the use state information and the stability information.

In the process, if the using state information accords with a first abnormal condition or the stability information accords with a second abnormal condition, the server determines that the target prop parameter is in an abnormal state; otherwise, the server determines that the target prop parameter is in a normal state.

Optionally, the first exception condition may be at least one of the average value being greater than the average value threshold or the standard deviation being greater than the standard deviation threshold, the second exception condition may be at least one of the distribution result not complying with the normal distribution or the degree of difference being greater than the target threshold, and the embodiment of the present application does not specifically limit the first exception condition or the second exception condition.

In some embodiments, after completing the anomaly detection of a target property parameter, the server may replace the target property parameter, and repeat the operations performed in step 202 and step 204 to detect whether the replaced target property parameter is abnormal or not, until all property parameters of the interactive property are traversed, and the anomaly detection of the entire property parameter of the interactive property is completed.

All the above optional technical solutions may be combined arbitrarily to form the optional embodiments of the present disclosure, and are not described herein again.

According to the method provided by the embodiment of the application, the target prop parameters of the interactive props are obtained through a plurality of test values in different using processes, the using state information and the stability information of the target prop parameters are quantized according to the test values, whether the target prop parameters are abnormal or not is detected according to the using state information and the stability information, the influence of user subjective factors in an abnormity detecting process can be avoided, the target prop parameters causing abnormity are accurately detected, and the application program providing the interactive props is optimized.

In the last embodiment, the server determines each test value of the target prop parameter based on each data item, so as to determine whether the target prop parameter is abnormal or not by performing data processing and abnormality detection on each test value. Since the data item may be in a text form or an image form, the embodiment of the present application takes the data item as an example of a target image, and details the process of acquiring the data set in step 201 described above.

Fig. 3 is a flowchart of acquiring a data set according to an embodiment of the present application, and referring to fig. 3, taking a data item as a target image as an example, the embodiment includes:

301. the server selects the interactive prop with the prop parameters configured by the configuration file in the application program.

In the above process, since the interactive prop usually has a plurality of prop parameters, and a plurality of sets of prop parameters can be formed by permutation and combination of different prop parameters, the user can configure the plurality of sets of prop parameters through the configuration file, and then the server selects the corresponding interactive prop in the application program according to the prop parameters configured by the configuration file.

For example, taking the configuration file as weipon config.xml as an example, test cases (test cases) can be managed by the configuration file, and one test case refers to a sample used in a test process. In the embodiment of the present application, it is assumed that the interactive property is a firearm-like interactive property, and the test case is used to indicate a firearm Identification (ID), a firearm name, and a firearm accessory ID used in each set of tests, where the firearm accessory may include a telescope, a muzzle, a grip, a stock, and the like. The configuration file can manage the test cases through the configuration table, for example, if two different firearm interaction objects, namely, the webon 1 and the webon 2 are selected, 3 different test cases can be configured, and the configuration statements are as follows:

in the configuration statement, the accessory of the webon 1 in the first test case selects only the number 1 times mirror, the accessory of the webon 1 in the second test case selects only the number 1 times mirror and the number 1 muzzle, the accessory of the webon 2 in the third test case selects only the number 1 times mirror, and the three test cases correspond to three different property parameter selections, so that three data sets can be generated in the subsequent steps.

302. The server determines a fixed shooting position and a shooting target for the virtual object in a virtual scene provided by the application program.

In the above process, the server may start an application program, display a virtual scene in the application program, and select a certain fixed shooting position and shooting target in the virtual scene. The shooting position refers to a position where the virtual object launches the interactive prop, and the shooting target refers to another virtual object to be shot by the virtual object, that is, the shooting target may be any virtual object, NPC, AI, or enemy virtual object except the virtual object in the virtual scene.

Through selecting fixed shooting position and shooting target, can carry out a lot of tests to interactive stage property under the condition of control variable, because shooting position, shooting target all remain unchanged when guaranteeing test at every turn, have just also guaranteed that the reason that causes the shooting effect to change is only for the stage property parameter of interactive stage property self to help getting rid of extra interference factor at the in-process that carries out the anomaly detection to target stage property parameter.

303. The server controls the virtual object to continuously shoot at the shooting position to the shooting target, and image capture is carried out on the virtual scene within the target duration to obtain a test image.

In the process, the server can control the virtual object to move to the shooting position, control the virtual object to aim at the shooting target at the shooting position, and use the interactive prop to aim at the target to continuously fire, so that the effect of continuous shooting is achieved.

Alternatively, if the interactive prop is itself in a single firing mode, then the interactive prop needs to be triggered to fire multiple times in succession, and if the interactive prop is itself in a multiple firing mode, then the interactive prop needs to be triggered to fire continuously only once.

Due to the influence of factors such as recoil, random variables, and the like, even in the case where the shooting position, the shooting target, and the sighting center are ensured to be unchanged, the positions at which bullets hitting the shooting target fall may be different from each other. When a bullet hits a shooting target, a bullet hole is usually left in the shooting target, and the bullet hole in a virtual scene does not permanently remain and fades or disappears with the lapse of time, so that the server needs to intercept an image of the virtual scene before one or more bullet holes disappear (that is, within the target duration) due to continuous shooting, and acquire the intercepted image as a test image. Where the target duration is any duration greater than or equal to 0, e.g., 5 seconds after the last bullet shot.

In some embodiments, the number of consecutive shots may be a fixed number that is greater than or equal to 1 and less than the loading amount, and of course, the number of consecutive shots may also be the loading amount, and at this time, the prop is directly controlled to continuously fire until all the loaded bullets in the clip are completely fired, and the firing may be stopped.

Fig. 4 is an interface schematic diagram of a test image according to an embodiment of the present application, and referring to fig. 4, a shooting target is determined as a shooting target 401 in a virtual scene 400, the number of continuous shots is determined as a loading amount, continuous shots are performed by controlling a virtual object 402 to aim at a target center of the shooting target 401, so that a series of shot holes are punched on the shooting target 401, and a screenshot of the virtual scene 400 within a target duration is obtained, so that the test image with each shot hole is captured.

304. The server repeatedly executes the step of controlling the virtual object to shoot continuously and obtain the test image in the step 303 to obtain a plurality of test images with the same item parameters, and determines the plurality of test images as an image set.

Step 304 is similar to step 303 and will not be described herein.

Based on the example of step 301, assuming that the steps of continuously shooting and capturing the test image are repeated 50 times under the prop parameter of the first test case, an image set which includes 50 test images and corresponds to the first test case can be obtained.

In the process, because the shooting behavior is random, the same set of property parameters are repeatedly sampled for many times, a test image in an image set can be intercepted for each set of property parameters for the subsequent calculation and analysis process, and the uncertainty caused by accidental errors can be reduced.

305. The server repeatedly executes the step 304 to obtain a plurality of image sets, wherein each image set comprises a plurality of test images generated when the interactive prop with the same prop parameter is used for multiple times.

Step 305 is similar to step 304, and is not described herein.

Based on the above example of step 304, the server repeatedly executes the step of obtaining the image sets for each test case, so as to obtain a plurality of image sets, where the number of the image sets is equal to the number of the test cases.

In step 301-.

Optionally, when storing the plurality of image sets, the server may name the plurality of image sets with respective timestamps and prop parameters. When the size of the test images is large, the test images can be randomly stored in different directories, and the time stamps and the prop parameters are used as key values due to the naming mode of the time stamps and the prop parameters, so that the directories for storing the test images are traversed, the test images with the same key values are divided into the same image set, and the image set can be conveniently and quickly divided.

In some embodiments, the server may also name the test images without using their respective timestamps and property parameters, but directly store the test images of the same image set into the same directory manually by the user, thereby also ensuring accurate division of the image set.

For example, the test images may be named in the form of "timestamp _ firearm ID _ scope ID _ muzzle ID _ grip ID _ stock ID", and the format of each test image is unified into jpg or png, which facilitates the subsequent classification and grouping of each test image. For example, if there is no accessory such as a telescope and a muzzle, the ID of the corresponding accessory may be set to 0.

Fig. 5 is a schematic diagram of test image naming provided in an embodiment of the present application, and referring to fig. 5, a plurality of test images with different timestamps can be obtained by repeatedly testing an interactive prop with a firearm ID of 001 and a scope ID of 1 and without configuring other accessories, and a naming example of a partial test image is shown on the left side. The server traverses the catalog of the test images by taking the 'firearm ID _ scope ID _ muzzle ID _ grip ID _ stock ID' as a key value, for example, the key values of the test images shown in the figure are '001 _1_0_0_ 0', and a series of test images with the same key value and different time stamps can be obtained.

306. The server acquires a target area and a filtering area corresponding to the plurality of image sets, wherein the filtering area is located in the target area.

The target area refers to an effective area for bullet hole extraction in the test image, and the filtering area refers to a noise area with shielding or interference in the target area. The target area and the filter area are determined based on the shot position and the shot target selected in step 302 above, and since each image set has the same shot position and shot target, each image set corresponds to the same target area and filter area.

In step 306, the server may issue any one of the plurality of image sets to the terminal, the user performs target area labeling and filtering area labeling on the test image on the terminal, the terminal sends the labeling result of the target area and the labeling result of the filtering area to the server, and the server analyzes the labeling result of the target area and the labeling result of the filtering area, so as to obtain the target area and the filtering area corresponding to each image set, thereby avoiding waste of human resources due to repeated labeling.

Fig. 6 is a schematic diagram of a target region and a filter region provided in an embodiment of the present application, referring to fig. 6, when a shooting target 401 in a virtual scene 400 is taken as a shooting target, a target surface region 601 of the shooting target 401 may be taken as the target region, and a region 602 in the target surface region 601 that is blocked by a head of a virtual object is taken as the filter region, it should be noted that fig. 6 only takes 1 filter region included in the target region as an example, and actually, the target region may include one or more filter regions, and the number of the filter regions is not specifically limited in the embodiment of the present application.

Fig. 7 is a schematic diagram of a target area and a filtering area provided in an embodiment of the present application, referring to fig. 7, when a wall 701 in a virtual scene 700 is taken as a shooting target, a middle part 702 of the wall may be taken as the target area, and two filtering areas are marked in the target area, which are an area 703 blocked by a head of the virtual object and an area 704 blocked by a box respectively. More or fewer filtering areas can be labeled according to the display and layout of each virtual article in the virtual scene, which is not described herein.

307. And the server performs target area cutting and gray level processing on each test image of the image sets to obtain a gray level image of the target area in each test image.

In the above process, for any test image, the server may cut out an image covered by the target area from the test image, and perform gray scale processing on the image covered by the target area, thereby obtaining a gray scale image of the target area.

When the gray scale processing is performed, any one of a component method, a maximum value method, an average value method, or a weighted average method may be used, for example, when the average value method is used, for each pixel point in an image covered by a target area, the server takes an average brightness value of each pixel point in R, G, B three channels as a gray scale value of each pixel point.

308. And the server carries out binarization processing on the gray level image of each target area, and sets the pixel points covered by the filtering area to be 0 in each binarized image to obtain a binary image of each target area.

In the process, the server can perform binarization processing according to a preset gray level threshold, and because the gray level value of a hit point (commonly called as a 'bullet hitting point', which means the position of a bullet hole on a shooting target after the bullet is shot by an interactive prop) of the bullet is lower, the pixel point with the gray level lower than the gray level threshold can be set to be 1, and the pixel point with the gray level higher than the gray level threshold is set to be 0, so that the binarization processing of the gray level image is completed. The gray threshold may be any value in the range of 0 to 255, for example, the gray threshold is set to 30.

Furthermore, because the filtering area is a marked noise area with shielding or interference, it can be considered that there is no impact point in the filtering area, all the pixel points in the filtering area are directly set to 0, so that the noise interference generated in the filtering area can be avoided, which is equivalent to performing primary noise reduction processing on the gray level image.

In step 306-308, the server extracts a target region from each test image of the plurality of image sets, and obtains a binary image of each target region. Because the impact points in the target area are distributed more densely, each impact point can be obviously displayed in the binary image of the target area, and the binary image of the target area can also be commonly referred to as a 'bullet hole extraction map'.

Fig. 8 is a schematic diagram of a bullet hole extraction diagram provided in an embodiment of the present application, and referring to fig. 8, the bullet hole extraction diagram shown in fig. 8 can be obtained by performing target area clipping, grayscale processing, and binarization processing on the basis of fig. 6, where in the bullet hole extraction diagram, the impact points are white, and the other pixel points except the impact points are all black.

In some embodiments, the server may also not obtain the labeled filter region, but only obtain the target region, so that after the binarization processing is performed on the grayscale image of the target region, a binary image can be directly obtained, and the flow of the anomaly detection process can be simplified.

309. And the server randomly selects at least one binary image from each image set to test, adjusts the processing parameters of the binarization processing process if the test result does not meet the target condition, and repeatedly executes the steps of obtaining the binary images and carrying out abnormity detection until the test result meets the target condition, wherein the target condition is used for representing the acceptable range of errors generated by the processing parameters.

Wherein the processing parameter comprises at least one of a target region, a filter region, or a gray level threshold.

Alternatively, the target condition may be that the extraction rate of the impact points is within an error range, the extraction rate of the impact points refers to a ratio of the impact points extracted from the binary image to the original impact points in the test image, and the error range of the extraction rate may be an interval. For example, the error range may be set to 95% to 105%, when the extraction rate is greater than 105%, it indicates that the server mistakenly considers many noise pixels as impact points, so that the accuracy of subsequent anomaly detection is affected, and when the extraction rate is less than 95%, it indicates that the server fails to extract the original impact points in the test image, so that the accuracy of anomaly detection is also affected.

In the process, by randomly selecting at least one binary image, small-batch testing can be performed on each image set, processing parameters adopted in the image processing process have higher adaptability, and the processing parameters most suitable for the current testing task can be determined by small-batch testing and iterative adjustment aiming at different testing tasks of different application programs, so that the processing parameters are optimized, and the accuracy of anomaly detection is improved.

In some embodiments, in iteratively adjusting the processing parameters, the server may follow any or at least one of the following parameter adjustment rules:

(1): when the impact points in the test image fall outside the binary image, expanding the target area until the missed impact points are covered in the target area;

(2): when a large number of non-bullet hole objects exist in the binary image and are extracted, selecting an area where the large number of non-bullet hole objects are gathered as a filtering area, for example, selecting the head of a virtual object in the target area of fig. 6 as the filtering area.

(3): when the impact point falls within the binary image but there is a small amount of omission, the gradation threshold value is increased stepwise, and for example, iterative adjustment may be performed by increasing the pixel value by 5 pixels at a time, while an upper limit may also be set for the gradation threshold value, and assuming that the gradation threshold value is represented by T, the upper limit of the gradation threshold value T may be set to max (T) ═ 70.

(4): when the impact point falls within the binary image but there is a small amount of noise, the grayscale threshold is gradually reduced, for example, iterative adjustment may be performed by reducing the value by 5 pixels each time, and a lower limit may also be set for the grayscale threshold, and assuming that the grayscale threshold is represented by T, the lower limit of the grayscale threshold T may be set to min (T) 10.

The server iteratively adjusts the parameters of at least one binary image in each image set until the test result meets the target condition, stops the iteration, and extracts the impact points of all the test images in all the image sets in batch by similar operations in the steps 307-308 on the basis of the processing parameters adopted in the last iteration to obtain each final binary image in each image set.

310. And the server performs anomaly detection on each binary image, deletes the anomaly image in each binary image to obtain each target image, and determines the target image corresponding to the same image set as one data set in the multiple data sets.

In the above process, the server screens the binary image finally generated in step 308, screens out abnormal images in the binary image, and deletes the abnormal images in the image set, so that each retained binary image is determined as a target image (i.e., a data item). The abnormal image may be a binary image in which no impact point is extracted, or a binary image in which the number of impact points exceeds a number threshold, for example, the number threshold may be twice the payload.

Fig. 9 is a schematic diagram of an abnormal image provided in an embodiment of the present application, and referring to fig. 9, after a virtual object is shot continuously due to network fluctuation generated when multiple tests are performed according to a configuration file, a shooting target has no pin hole, a binary image (pin hole extraction map) generated according to the abnormal image also belongs to an invalid abnormal image, and a server deletes the abnormal image from an image set after detecting the abnormal image.

Due to the fact that network fluctuation or shooting failure may occur when a test image is generated, some abnormal images are mixed in the test image, whether the binary image is an invalid abnormal image can be finely screened through the screening process of the binary image, if a certain binary image is determined to be the abnormal image, the test image corresponding to the binary image can also be directly determined to be the abnormal image, the abnormal image is deleted, the abnormal image is prevented from being included in a sample range of follow-up statistical analysis, the automatic detection and filtering effects of the abnormal image are achieved, and accuracy and intelligence of the abnormal detection process are improved.

The method provided by the embodiment of the application takes the data item as the target image as an example, tests the interactive prop for multiple times through the configuration file to generate a plurality of image sets, by cutting the target area, processing the gray scale and binaryzation for each test image in each image set, a plurality of binary images are obtained, the binary images can reflect the impact points extracted from the test image, and the processing parameters adopted in the image processing can be more suitable for the current abnormality detection task by iteratively adjusting the processing parameters, detecting abnormality of each binary image to detect abnormal images in the binary image, deleting the abnormal images to obtain multiple target images, and each target image is grouped to obtain each data group, so that random errors caused by abnormal images can be avoided, and the accuracy of the whole abnormal detection process aiming at the target prop parameters is improved.

In the embodiment of the present application, after the data group is obtained, taking a parameter of the target prop as a maximum offset between a shooting position and a target position of the interactive prop as an example, how to perform data processing and statistical analysis on the maximum offset is described, and since the maximum offset is positively correlated with the recoil, the use state information and the stability information of the recoil can be reflected by the use state information and the stability information of the maximum offset.

Fig. 10 is a flowchart for analyzing the maximum offset according to an embodiment of the present application, and referring to fig. 10, the embodiment includes:

1001. and if the target prop parameter is the maximum offset between the shooting position and the target position of the interactive prop, the server selects a reference point in any target image.

Wherein the target position is also the impact point of the bullet.

Alternatively, the reference point may be the geometric center of the test image corresponding to the target image, in which case the collimation point at which the virtual object is aimed is always located at the geometric center of the test image, and the picture displayed by the application program automatically returns to the position at the time of the first aiming shot after each shooting.

Optionally, the reference point may also be a pixel point with a pixel value of 1 located at the lowest position in the target image, and since the pixel point with the pixel value of 1 is also an extracted impact point in the target image, in other words, the impact point located at the lowest position may be used as the reference point, in this case, a picture displayed by the application program after each shooting does not automatically return to a position at which the shooting is aimed for the first time, but due to the influence of recoil, an offset of the impact point of each bullet in the upward direction is gradually increased during continuous shooting, and thus the impact point located at the lowest position is used as the reference point.

Fig. 11 is a schematic diagram of a selected fiducial provided by an embodiment of the present application, and referring to fig. 11, the server directly determines the geometric center of the test image as the fiducial. Fig. 12 is a schematic diagram of a selected reference point provided in an embodiment of the present application, and referring to fig. 12, the server directly uses the lowest impact point as the reference point. It should be noted that if the impact point located at the bottom coincides with the geometric center of the test image, the two ways of selecting the reference point are consistent.

1002. And the server determines at least one target pixel point which is farthest from the reference point in each direction from the pixel points with the pixel values of 1 in the target image.

After the reference point is determined, distances in all directions between each pixel point (that is, the impact point) with a pixel value of 1 and the reference point may be determined, and a target pixel point farthest from each direction may be determined, for example, target pixel points in four directions, i.e., an upper direction, a lower direction, a left direction and a right direction, may be selected, of course, the selected direction may be any direction, for example, a direction of 45 degrees north-east, and the like, and the embodiment of the present application does not specifically limit the selected direction.

1003. And the server determines the distance between the at least one target pixel point and the reference point in each direction as the maximum offset in each direction, and determines the maximum offset in each direction corresponding to the target image as a test value of the target prop parameter.

In the above process, for any direction, the server determines the distance between the corresponding target pixel point and the reference point in the direction as the maximum offset of the direction, and repeats the above steps for each direction, so as to obtain the maximum offsets of the directions, and the maximum offsets can form a test value sequence.

In the above-mentioned step 1001-1003, it is illustrated how the server implements the operation of step 202 in the above-mentioned embodiment when the data item is the target image and the target item parameter is the maximum offset between the shooting position and the target position, that is, the server determines the plurality of test values of the target item parameter according to each data item of each data set, wherein the test values are represented in the form of a test value sequence formed by a plurality of maximum offsets.

Fig. 13 is a schematic diagram of an upward maximum offset provided in an embodiment of the present application, and referring to fig. 13, in a target image, a target pixel point of a reference point a that is farthest in an upward direction is a point B, and a distance between the reference point a and the target pixel point B in a vertical direction is taken as the upward maximum offset of the target image.

1004. And the server acquires at least one of the average value or the standard deviation of each test value, and determines at least one of the average value or the standard deviation as the use state information of the target prop parameter.

In the above process, the usage state information is used to reflect an average usage status of the target prop parameter in multiple rounds of usage processes, for example, the usage state information may include an average value and a standard deviation of each test value, the average value is used to measure an average size of the test values, the average size of the test values is larger when the average value is larger, the standard deviation is used to measure a dispersion degree of each test value, and the randomness of the target prop parameter is larger when the standard deviation is larger, because some random variables may be introduced in the usage process of the interactive prop, the randomness of the target prop parameter may be measured by the standard deviation.

For example, the server generates 50 target images for each test case as a data group, so that the average value and the standard deviation are used as analysis indexes of random variables, the average value and the standard deviation of the maximum offset of the 50 target images in each direction in each data group are respectively calculated, the average value reflects the recoil of the interactive prop, the recoil of the interactive prop is stronger when the average value is larger, the standard deviation reflects the randomness of the interactive prop, and the higher the standard deviation is, the higher the randomness of the interactive prop exists when the interactive prop is shot. Of course, the number of target images included in the data set is not limited to 50, the number of target images may be any value greater than or equal to 1, and when the number of target images is larger, the more abundant the description sample is, the more accurate the analyzed statistical result is.

1005. The server acquires a plurality of comparison groups corresponding to a plurality of data groups in which the test values are located.

Since the server determines each test value according to each data item, each test value may also have a grouping corresponding to each data item, that is, each test value corresponds to a different data group, and therefore, the server may calculate an average value and a standard deviation of the test values in each data group by using the data group as a unit.

Furthermore, any two different data sets can be mutually compared, and when any data set is processed, one comparison set is selected for the data set, so that the data set and the corresponding comparison set can be subjected to statistical analysis, and stability information of the data set can be conveniently obtained. Wherein the stability information may include at least one of a distribution result or a degree of difference.

1006. And the server performs replacement inspection on the plurality of data groups and the plurality of comparison groups to obtain a distribution result of the target prop parameters.

For any one data set and the comparison set of the data sets, the server may perform the permutation check by performing the following sub-steps:

1006A, the server obtains a target mean value difference of the test values of the target prop parameters between any one of the data sets and the comparison set of the data set.

The data set currently selected for the permutation test may also be referred to as a "baseline set".

The permutation test (membership test) is a method for testing whether a certain hypothesis is true, and the hypothesis and the data of the same test case in the comparison group are stable or distributed, so as to test whether the hypothesis is true in the process of repeated calculation.

For example, assuming that the data of the same test case in the data group and the comparison group are stable or uniformly distributed, taking any test case as an example for explanation, it is assumed that the upward directions in the data group of the test case and the comparison group respectively include 50 maximum offsets. The average of the 50 maximum offsets within the data set is identified by t1 and the average of the 50 maximum offsets within the comparison set is represented by t 2. The target mean difference t0 between the data set and the comparison set can be expressed as t 0-t 1-t 2.

1006B, the server randomly divides the test values of the data set and the comparison set into two different permutation sets, and determines the average difference between the two permutation sets.

In the above process, the data group and the comparison group contain 100 maximum offsets in total, and the server may combine the 100 maximum offsets into one sample set, randomly divide all the maximum offsets into two permutation groups, each of which contains 50 maximum offsets, and determine the mean difference between two permutation groups in a similar manner to step 1006A.

1006C, the server repeatedly executes the steps of randomly dividing the permutation group and determining the mean value difference in the step 1006B for multiple times to obtain multiple mean value differences, and obtains a first sample distribution formed by the multiple mean value differences.

In the above process, after the step 1006B is repeatedly executed for multiple times and multiple mean differences are obtained, the server may perform data fitting on the multiple mean differences, so as to obtain a first sample distribution.

For example, 10000 mean differences can be obtained by repeating the above step 1006B 10000 times, and a mean difference sampling distribution map can be generated by performing data fitting on the 10000 mean differences, and the first sampling distribution to which the mean differences obey can be reflected by the mean difference sampling distribution map.

Fig. 14 is a schematic diagram of a first sampling distribution provided in an embodiment of the present application, and referring to fig. 14, a frequency distribution histogram of mean differences is selected as a mean difference sampling distribution graph, a horizontal axis is used to represent the mean differences, and a vertical axis is used to represent distribution proportions (i.e., probabilities occurring in 10000 times of repeated experiments) corresponding to the mean differences, and a probability curve of the fitted first sampling distribution is drawn in the frequency distribution histogram.

1006D, the server obtains a probability value of the permutation test according to the confidence interval of the target mean value difference falling in the first sampling distribution, and determines the distribution result of the target prop parameter based on the probability value.

In the above process, the server may detect whether the target mean difference t0 falls within a predetermined confidence interval in the first sample distribution, and determine whether the assumption is true by acquiring a probability value (P value). For example, the confidence interval may be chosen to be the interval in which 99% confidence is present.

For example, assuming that the predetermined confidence interval is an interval with a 99% confidence, if the statistics (i.e., mean difference) greater than | t0| in 10000 iterations has n (n ≧ 0), the probability value P may be obtained as P ═ n/10000, so that the smaller n is, the smaller the number of occurrences of the target mean difference t0 in 10000 iterations is, the smaller the probability of occurrence of the corresponding t0 is, and the smaller P is. Whereas if the P value is less than 1%, indicating that the number of occurrences of the target mean difference t0 is small in 10000 replicates, the assumption that the data between the data set and the comparison set is stable or homographic is not established, because if the assumption is established, indicating that the target mean difference t0 has a greater probability of occurrence, the P value cannot be less than 1%.

In the embodiment of the present application, 10000 times of repeated tests in the permutation test are taken as an example for explanation, and the number of times of repeated tests may be any value greater than or equal to 1, and the number of times of repeated tests in the permutation test is not specifically limited in the embodiment of the present application. In one possible implementation, a minimum of 2000 replicates may be performed.

In some embodiments, the server may not perform the above step 1006, but directly perform the following step 1008 after performing the following step 1007, and determine the degree of difference as the stability information, which can simplify the flow of abnormality detection.

1007. And the server carries out self-service statistics on the plurality of data groups and the plurality of comparison groups to obtain the difference degree of the target prop parameters.

In the above step 1006, the permutation test is performed on each data group only for one comparison group, that is, t0 is calculated only once for the data group and the comparison group, however, due to the randomness of each test value of the comparison group itself, the P value obtained by the permutation test is not reliable enough, so that the reliability of the P value in the permutation test can be tested by adding statistics through self-service statistics (bootstrap) in the above step 1007, which will be described in detail below.

For any data set and the comparison set corresponding to the data set, the server can perform the replacement check by executing the following sub-steps:

1007A, the server acquires the probability value of any data group and the comparison group corresponding to the data group in the replacement test.

Step 1007A is similar to step 1006, and is not described herein. That is, the self-service statistics can be performed only on the basis of acquiring the P value by performing replacement check on any data group and the comparison group.

1007B, the server randomly extracts the test values from the comparison group to form a test group, and performs replacement test based on the data group and the test group to obtain a test probability value.

In the above process, it is necessary to keep the data set unchanged, and repeat the random extraction with replacement for the comparison set, for example, the comparison set contains 50 maximum offsets, and the server repeats the random extraction with replacement 50 times, so as to form a test set. A trial probability value may be obtained by performing a permutation test operation similar to step 1006 above on the basis of the data set and trial set.

1007C, the server repeatedly executes the steps of forming the test group and obtaining the test probability value in the step 1007B to obtain a plurality of test probability values, and obtains a second sample distribution formed by the plurality of test probability values.

In the above process, after the step 1007B is repeatedly performed for multiple times, multiple experimental probability values may be obtained, so as to perform data fitting on the multiple experimental probability values to obtain a second sample distribution.

For example, 10000 times of repeating the above step 1007B may obtain 10000 experimental probability values, and performing data fitting on the 10000 experimental probability values may generate a probability value sampling distribution map, where the probability value sampling distribution map may reflect the second sampling distribution to which the experimental probability value is obeyed.

Fig. 15 is a schematic diagram of a second sampling distribution provided in the embodiment of the present application, and referring to fig. 15, a dot diagram of the test probability value is selected as a probability value sampling distribution diagram, a horizontal axis is used for representing the test probability value, and a density degree of data points in each data interval is used for reflecting an occurrence probability of the test probability value in the data interval.

Fig. 16 is a partial enlarged view of a second sampling distribution provided in the embodiment of the present application, and referring to fig. 16, data points in an interval of 0 to 0.01 in fig. 15 are enlarged and displayed, it can be seen that when a P value (test probability value) is less than 0.01, only 3 data points are included, that is, in 10000 times of repeated tests, only 3 test probability values are less than 0.01.

1007D, the server obtains the difference degree of the target prop parameter according to the occurrence frequency of the designated interval in the second sampling distribution of the probability value.

In the above process, the specified interval may be a predetermined confidence interval, for example, the specified interval may be selected as P < 0.01. In an exemplary scenario, as analyzed with respect to fig. 16, the number of times P <0.01 in 10000 repeated experiments is 3, that is, the confidence of P <0.01 is 3/10000 ═ 0.03%, so that it is proved that the probability of occurrence of difference between the data set and the comparison set is extremely low, and thus the confidence that the data set and the comparison set obey the same distribution is considered to be high, that is, it is determined that the data set and the comparison set really obey the same distribution.

Fig. 17 is a schematic diagram of a second sample distribution provided in an embodiment of the present application, and fig. 17 is a probability value sample distribution diagram represented by a dotted graph, similar to fig. 15, but using different data groups and contrast groups, so that the two reflect the different second sample distribution.

Fig. 18 is a partial enlarged view of a second sampling distribution provided in the embodiment of the present application, and referring to fig. 18, data points within an interval of 0 to 0.01 in fig. 17 are enlarged and displayed, it can be seen that when a P value (test probability value) is less than 0.01, 6798 data points are included, that is, in 10000 times of repeated tests, 6798 test probability values are less than 0.01. That is, the number of times of P <0.01 in 10000 replicates was 6798, and the confidence of P <0.01 was 6798/10000-67.98%, indicating that the confidence that the data set and the comparison set obeyed the same distribution was low, i.e., it was determined that there was a high difference between the data set and the comparison set.

In the embodiment of the present application, 10000 times of repeated execution in the self-service statistics are taken as an example for explanation, and the number of times of repeated tests may be any value greater than or equal to 1, and the number of times of repeated tests in the self-service statistics is not specifically limited in the embodiment of the present application. In one possible implementation, a minimum of 2000 replicates may be performed.

In some embodiments, the server may not perform step 1007, but directly perform step 1008 after performing step 1006, and determine the distribution result as the stability information, which can simplify the flow of anomaly detection.

1008. And the server determines at least one of the distribution result or the difference degree as the stability information of the target prop parameter.

In the aforementioned step 1006-1008, it is equivalent to the server performing statistical analysis on the plurality of data sets and the plurality of comparison sets to obtain the stability information of the target property parameter. In the embodiment of the present application, how to obtain the distribution result and the difference degree is respectively illustrated by using a single data group and a single comparison group as an example, so that similar operations are repeatedly performed on each data group and the corresponding comparison group, the distribution result and the difference degree of each of the plurality of data groups and the plurality of comparison groups can be obtained, and at least one of the two is determined as the stability information.

Through the

above steps

1004 and 1008, the server realizes the operation executed in the step 203 in the above embodiment, that is, the server determines the use state information and the stability information of the target item parameter according to each test value of the target item parameter.

1009. And the server detects whether the target prop parameter of the interactive prop is abnormal or not according to the use state information and the stability information.

Step 1009 is similar to step 204, and is not described herein.

Based on the above example, if the average is greater than a certain offset threshold, it may be determined that the maximum offset is greater, and therefore the squat force is determined to be abnormal (higher), or if the standard deviation is greater than a certain standard deviation threshold, it may be determined that the randomness of the maximum offset is greater, and therefore the squat force is determined to be abnormal (higher randomness), or if the distribution results in that the data set and the contrast set do not follow the same distribution, it may indicate that the maximum offset is unstable between the different data sets, and therefore the squat force is determined to be abnormal (lower stability), or if the confidence that the data set and the contrast set follow the same distribution is low, it may also be considered that the data set and the contrast set do not follow the same distribution, and it may indicate that the maximum offset is unstable between the different data sets, and therefore the squat force is determined to be abnormal (.

It should be noted that, in the embodiment of the present application, only the target prop parameter is taken as an example to exemplarily describe the flow of abnormality detection, optionally, the target prop parameter may also be a random variable, different firearm accessories are added for the interactive prop, and the like, and similar steps may be performed to perform abnormality detection, and the content of the target prop parameter is not specifically limited in the embodiment of the present application.

The method provided by the embodiment of the application takes a target prop parameter as an example, obtains a test value corresponding to each target image by selecting a reference point and determining a target pixel point for each target image, obtains an average value and a standard deviation of each test value, obtains use state information of the target prop parameter, can measure the recoil strength and the randomness of the interactive prop by using the state information, obtains a data group where each test value is located and a corresponding comparison group, respectively carries out replacement check and self-help statistics based on the data group and the comparison group, can avoid system errors caused by the conditions of limited sample number, limited comparison group number and the like, obtains the distribution result and the difference degree of the target prop parameter, can measure whether the recoil strength of the interactive prop is stable in the process of multiple uses by using the distribution result and the difference degree, therefore, when abnormity detection is carried out based on the use state information and the stability information, whether the recoil force of the interactive prop is abnormal or not can be determined.

Based on the various embodiments described above, in one exemplary scenario, a method for detecting whether an abnormality occurs in a sense of a manipulator in an FPS game is provided. The firearm hand feeling refers to a general term of user experience brought by various property parameters of the firearm property in the FPS game.

The server may provide a numerical analysis platform for the test terminal, where the numerical analysis platform may be a website or an application program, and the embodiment of the present application does not specifically limit the form of the numerical analysis platform. The user can log in the numerical analysis platform through the test terminal, and whether the firearm handfeel of various FPS games is abnormal or not can be detected through the numerical analysis platform after the user successfully logs in the numerical analysis platform.

Taking a numerical analysis platform as a website as an example, a user logs in the website on a test terminal, clicks a firearm hand feeling analysis option in the website, and can enter a management page for firearm hand feeling analysis, a test task can be created and managed in the management page, after the test task is created, the user can set parameters such as an FPS game, a target prop parameter, the number of times of replacement inspection, the number of times of self-service statistics and the like of the test task, and the user triggers the test terminal to send a test instruction to a server, wherein the test instruction carries the parameters. When the server receives the test instruction, each parameter in the abnormality detection process is set according to the test instruction, automatic abnormality detection is carried out on the FPS game, and after the detection is finished, a test result is returned to the test terminal.

Optionally, the test result may at least carry an "abnormal" or "normal" status result, and optionally, the test result may also carry at least one of each generated test image, each generated binary image, each generated mean difference sample distribution map, or each generated probability value sample distribution map.

Fig. 19 is a schematic diagram of a firearm hand feeling analysis method provided in an embodiment of the present application, referring to fig. 19, a user logs in a numerical analysis platform at a test terminal, and after selecting a firearm hand feeling analysis option, the user triggers a test instruction, so that a server performs a test in response to the test instruction, fig. 19 also shows a test image and a corresponding binary image, where the test image and the binary image may be included in a test result and sent to the test terminal by the server after the detection is completed, and of course, the test image and the binary image may also be stored in the server, and the user may select whether to download to the test terminal.

Fig. 20 is a schematic diagram of a firearm hand feeling analysis system provided in an embodiment of the present application, and referring to fig. 20, in the above exemplary scenario, when analyzing and detecting a firearm hand feeling through an interaction process between a test terminal and a server, the data module 2001, the calculation module 2002, and the result presentation module 2003 may be logically divided. Typically, the data module 2001 and the calculation module 2002 are deployed on the server side, and the result presentation module 2003 is deployed on the user side (i.e., the test terminal).

The data module 2001 is used for generating image sets and preprocessing data, and can generate a plurality of test images from a data source, group each test image to obtain a plurality of image sets, and preprocess the plurality of image sets to obtain a plurality of data sets.

It should be noted that, when the data module 2001 generates the image set from the data source, the type of the data source may include image data generated by a configuration file or game data collected at the bottom layer, and the embodiment of the present application does not specifically limit the configuration of the data source. The image data and the game data of a plurality of types of FPS games are usually stored in the data source, so that when the handfeel of the firearm is analyzed, the handfeel of the firearm is not limited to analyzing the handfeel of different firearm props (for example, firearms with different models and the same accessories, firearms with different accessories and the same models) in the same type of FPS game, but also the handfeel of the firearm props in different FPS games can be contrastively analyzed, and the embodiment of the application is not particularly limited thereto.

The calculation module 2002 is configured to perform core operations such as cartridge hole extraction based on image processing (generating a cartridge hole extraction map), abnormal image detection (iteratively adjusting parameters and detecting an abnormal image), test value calculation of a target prop parameter (for example, calculating a jitter offset), and stability detection based on statistical analysis (replacement testing and self-service statistics).

Optionally, the calculating module 2002 may extract features (for example, bullet hole position features) from the test image so as to analyze and calculate the test value of the current target prop parameter, or may directly analyze the test value of the current target prop parameter from the underlying game data, and the embodiment of the present application does not specifically limit the obtaining manner of the test value. A complex and complete data analysis can be performed by the calculation module 2002 based on the respective test values.

The result display module 2003 is used for displaying abnormal data (abnormal image), a valid data bullet hole map (target image), a statistical analysis result (firearm monitoring result table generated based on stability information), and the like. The result presentation module 2003 may output the analysis result of the firearm feel in a suitable manner (image, text, table, or the like).

On the basis of the firearm hand feeling analysis system shown in fig. 20, in order to describe the functions of each module in detail, please refer to fig. 21, fig. 21 is a schematic diagram of a firearm hand feeling analysis system provided in an embodiment of the present application, and the functions of each module in fig. 20 are detailed in fig. 21, so as to better describe the operations performed by each module. The data module 2001 acquires an image set from a data source, manages the image set by a profile, preprocesses the image set to group images, and obtains a plurality of data groups. The calculation module 2002 extracts the bullet holes, generates a binary image of each test image, randomly extracts a part of the binary image to perform a small batch test, iteratively adjusts processing parameters when the binary image is generated to obtain each finally output bullet hole extraction image, performs abnormal data detection on each bullet hole extraction image, deletes abnormal images, determines the retained bullet hole extraction image as a target image, performs jitter offset calculation (that is, calculates the maximum offset in each direction) based on the target image, and performs statistical analysis based on the jitter offset to obtain a statistical analysis result. Further, when the computing module 2002 detects abnormal data, the detected abnormal image may form an abnormal data result, the detected retained target image may form a bullet hole extraction image, the computing module 2002 outputs the abnormal data result and the bullet hole extraction image to the result display module 2003 respectively, and then the invalid abnormal data result may be deleted, and in addition, the computing module 2002 may also output the statistical analysis result to the result display module 2003, and the result display module 2003 displays the abnormal data result, the bullet hole extraction image, and the statistical analysis result respectively.

In some embodiments, the calculation module 2002 may output the firearm monitoring result in a table form, as shown in table 1, after performing anomaly detection on3 test cases, the anomaly probability of each firearm prop can be obtained, where the anomaly probability of the firearm 3 is the highest and is up to 67.98%. Therefore, technicians can optimize the firearms 3, the optimization process of the FPS game can be improved, and the user experience provided by the FPS game is improved.

TABLE 1

Firearms prop	Probability of anomaly
		Weapon1 (firearm 1)	0.31％
Weapon2 (firearm 2)	0.09％
		Weapon3 (firearm 3)	67.98％

In addition, since the calculation module 2002 automatically detects and filters out possible abnormal data results, but still displays the abnormal data results in the result display module 2003, the user can conveniently check the abnormal data results at any time, thereby judging whether an abnormal image is misjudged. The user can also conveniently check the shot hole extraction image based on the result display module 2003, can randomly check the specific shot impact distribution condition, and is convenient for the user to manually analyze the abnormity and troubleshoot the fault. For the statistical analysis result, the firearm monitoring result table in the above table may be used for displaying, or other display manners such as a bar chart, a sector chart, a pie chart, and the like may be used for displaying, and an interface for customizing the statistical analysis result may be provided for the user, so as to support the user to customize the form and content of the statistical analysis result, for example, the user may customize the detailed data of each image, the statistical data of each data group, or only the result of firearm comparison monitoring, and the embodiment of the present application does not specifically limit the content and form of the statistical analysis result.

Fig. 22 is a schematic diagram of a firearm hand feeling analysis method provided in an embodiment of the present application, and referring to fig. 22, schematically, based on the firearm hand feeling analysis system, a data module 2001 acquires a plurality of test images from a data source, pre-processes and groups the plurality of test images to obtain a plurality of image sets, performs region labeling on data in the image sets to mark out a target region and a filter region, so as to perform bullet hole extraction by a calculation module 2002 to obtain each binary image, performs abnormal data detection on each binary image, can detect invalid data (abnormal image) in the binary image, obtains a retained target image after deleting the abnormal image, calculates a maximum offset of an impact point based on the target image, performs statistical analysis based on a permutation test and a self-help method to obtain a mean difference distribution map and a probability value sampling distribution map, the statistical analysis result (e.g., a shooter monitoring result table) is finally output to the result presentation module 2003.

Based on the firearm hand feeling analysis system provided by the embodiment of the application, different data sets and comparison sets are selected, so that the firearm hand feeling analysis system can be suitable for various test tasks, such as: whether the analysis judges firearms stage property reasonable in design, whether satisfy the design principle of high-low performance firearms stage property, whether the design of firearms stage property has the discrimination, whether the accessory of detecting firearms stage property takes effect, whether detect firearms stage property stable etc. in the different versions of playing.

Through configuring different test tasks, multidimensional abnormity detection can be carried out on the firearm hand feeling of the FPS game, and the following detailed description is carried out on several different test tasks:

firstly, testing the difference of the design performance of firearms props

When carrying out firearms stage property design in FPS recreation, should be better than the performance of naked rifle (referring to the firearms stage property of not installing any accessory) after the installation accessory with the firearms stage property, through the average value of the maximum offset of the firearms stage property of the different accessories of contrast installation, if the average value of the firearms stage property of installation accessory is greater than the average value of naked rifle, show that the recoil has increased on the contrary after the installation accessory, this unsatisfied firearms design principle, need adjust the parameter of firearms stage property, thereby can assist game testing personnel and developer to carry out firearms design inspection.

For another example, in the original design, the recoil of the high-performance firearm prop is small, and by comparing the average values of the maximum offsets of the firearm props with different performances, if the average value of the high-performance firearm prop is larger than the average value of the low-performance firearm prop, it is indicated that the high-performance firearm prop has a larger recoil on the contrary, which also does not satisfy the firearm design principle, and the parameters of the firearm prop need to be adjusted. Different firearm design principles can be configured with different data sets and comparison sets, which are not described in detail herein.

Distinguishing and detecting different firearm props and accessories

Different firearm props with the same accessories are selected to be installed, or the same firearm props with the different accessories are selected to be installed, and the same firearm props with the different accessories are respectively used as a reference group and a comparison group to be subjected to statistical analysis, so that whether obvious differences exist can be detected, whether the firearm accessories fail can be automatically detected in the above mode, problems (bugs) related to the accessories can be found in time, the efficiency of anomaly detection is improved, and the speed of anomaly investigation is improved.

Thirdly, testing the stability of the hand feeling of firearms in the same game version

With the same version of the FPS game, the firearm feel should be stable during the user's experience (i.e., use). At the moment, a certain firearm prop can be selected to carry out multi-round testing, the multi-round testing results are contrastively analyzed, if the confidence coefficient that P is less than 0.01 is small, no obvious difference exists, namely the firearm prop is stable in hand feeling and normal in performance, otherwise, the firearm prop is worth paying attention, and related anomalies need to be further checked.

Fourthly, detecting stability of firearm hand feeling among different game versions

By utilizing the analysis system, stable game version data can be used as monitoring data (data set for replacement inspection), a comparison experiment is carried out after a game new version is updated and before the game new version is released, the game data of the new version is used as a comparison set for replacement inspection, so that whether the handfeel of a firearm prop is stable between different game versions can be detected, whether the version updating has adverse effects on the handfeel of the firearm is detected, whether problems occur in the handfeel of the firearm can be early warned by detecting the difference of the handfeel of the firearm between two game versions, the problems are probably not caused by directly modifying the firearm data, and also can be caused by other functions, and bugs possibly existing under more conditions can be found by monitoring in the analysis system.

Fifthly, designing and adjusting property parameters of firearms property

The performance of the same firearm prop in different FPS games is contrastively tested, and the difference of the firearm prop can be quantitatively analyzed, so that the relevant prop parameters with more reasonable hand feeling attributes can be adjusted, and the FPS game with better firearm hand feeling experience is designed.

The firearm hand feeling analysis system provided by the embodiment of the application solves the problem of game data loss (due to reasons such as confidentiality) based on an image processing method, and solves the problem of difficulty in random variable measurement and analysis aiming at a non-parameter difference detection method (displacement test and self-help method statistics) of significance in random distribution statistics. Taking analysis of firearm recoil as an example, a set of general firearm hand feeling analysis scheme is established, can be applied to measurement and analysis of other hand feeling attributes, and provides a direction and feasible scheme for quantitative analysis of the hand feeling of the FPS game firearm.

Fig. 23 is a schematic structural diagram of an abnormality detection apparatus provided in an embodiment of the present application, and referring to fig. 23, the apparatus includes:

an obtaining module 2301, configured to obtain multiple test values of a target item parameter of an interactive item in different use processes, where the target item parameter is used to reflect a target interaction characteristic of the interactive item in the use processes;

a determining module 2302, configured to determine, according to each test value of the target prop parameter, use state information and stability information of the target prop parameter;

a detecting module 2303, configured to detect whether the target property parameter of the interactive property is abnormal according to the usage state information and the stability information.

The device that this application embodiment provided, through a plurality of test values of the target stage property parameter of acquireing interactive stage property in different use, use state information and the stability information of target stage property parameter are quantifys according to each test value, thereby according to use state information and stability information, whether detect target stage property parameter and take place unusually, can avoid the influence of user's subjective factor in the unusual detection procedure, the accurate target stage property parameter that causes unusually that detects out, so that optimize the application program that provides interactive stage property.

In a possible implementation manner, based on the apparatus composition of fig. 23, the obtaining module 2301 includes:

and the determining unit is used for determining the plurality of test values of the target prop parameter according to the data items of the data groups.

In a possible implementation, the data item is a target image, and based on the apparatus composition of fig. 23, the first obtaining unit includes:

the system comprises a testing subunit, a processing unit and a processing unit, wherein the testing subunit is used for testing interactive props with different prop parameters for multiple times through a configuration file to obtain a plurality of image sets, and each image set comprises a plurality of testing images generated when the interactive props with the same prop parameters are used for multiple times;

the acquisition subunit is used for extracting target areas from each test image of the plurality of image sets and acquiring binary images of each target area;

In one possible embodiment, the obtaining subunit is configured to:

acquiring a target area and a filtering area corresponding to the plurality of image sets, wherein the filtering area is positioned in the target area;

In one possible embodiment, the apparatus is further configured to:

In one possible embodiment, the determining unit is configured to:

In one possible embodiment, the reference point is a geometric center of the test image corresponding to the target image; or, the reference point is a pixel point of which the lowest pixel value is 1 in the target image.

In one possible embodiment, the test subunit is configured to:

and repeatedly executing the steps of controlling the virtual object to shoot continuously and obtaining the test images to obtain a plurality of test images with the same prop parameters, and determining the plurality of test images as an image set.

In one possible embodiment, the plurality of image sets are named with respective timestamps and prop parameters when stored.

In one possible embodiment, based on the apparatus components of fig. 23, the determining module 2302 comprises:

the second acquisition unit is used for acquiring at least one of the average value or the standard deviation of each test value and determining at least one of the average value or the standard deviation as the use state information of the target prop parameter;

In a possible embodiment, based on the device composition of fig. 23, the analysis unit comprises:

the self-service method statistics subunit is used for carrying out self-service method statistics on the plurality of data groups and the plurality of comparison groups to obtain the difference degree of the target prop parameter;

In one possible embodiment, the permutation check subunit is configured to:

for any data group and a comparison group of the data group, acquiring a target average value difference of the test values of the target prop parameters between the data group and the comparison group;

repeating the steps of randomly dividing the permutation group and determining the average value difference for multiple times to obtain multiple average value differences and obtain a first sampling distribution formed by the multiple average value differences;

and obtaining a probability value of the permutation test according to a confidence interval of the target mean value difference falling in the first sampling distribution, and determining a distribution result of the target prop parameter based on the probability value.

for any data group and a comparison group corresponding to the data group, obtaining the probability value of the data group and the comparison group in the replacement test;

randomly extracting test values from the comparison group to form a test group, and performing replacement test based on the data group and the test group to obtain a test probability value;

repeatedly executing the steps of forming a test group and obtaining test probability values to obtain a plurality of test probability values, and obtaining second sampling distribution formed by the plurality of test probability values;

It should be noted that: in the foregoing embodiment, when detecting an abnormality, the abnormality detection apparatus is described by taking only the division of the functional modules as an example, and in practical applications, the functions may be distributed by different functional modules as needed, that is, the internal structure of the computer device may be divided into different functional modules to complete all or part of the functions described above. In addition, the anomaly detection device and the anomaly detection method provided by the above embodiments belong to the same concept, and specific implementation processes thereof are described in detail in the anomaly detection method embodiments, and are not described herein again.

Fig. 24 is a schematic structural diagram of a computer device according to an embodiment of the present application, where the computer device 2400 may be a server according to the foregoing embodiments, and the computer device 2400 may generate a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) 2401 and one or more memories 2402, where the memory 2402 stores at least one program code, and the at least one program code is loaded and executed by the processor 2401 to implement the abnormality detection method according to the foregoing embodiments. Of course, the computer device 2400 may also have components such as a wired or wireless network interface, a keyboard, and an input/output interface, so as to perform input and output, and the computer device 2400 may also include other components for implementing device functions, which are not described herein again.

In an exemplary embodiment, there is also provided a computer readable storage medium, such as a memory, including at least one program code, which is executable by a processor in a terminal to perform the anomaly detection method in the above embodiments. For example, the computer-readable storage medium may be a ROM (Read-Only Memory), a RAM (Random-Access Memory), a CD-ROM (Compact Disc Read-Only Memory), a magnetic tape, a floppy disk, an optical data storage device, and the like.

It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

The above description is only exemplary of the present application and should not be taken as limiting, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An anomaly detection method, characterized in that it comprises:

obtaining a plurality of test values of target prop parameters of the interactive prop in different using processes, wherein the target prop parameters are used for reflecting target interactive characteristics of the interactive prop in the using process, the plurality of test values are determined based on a plurality of target images, and the target images are images reserved after deleting abnormal images in an image set generated in the different using processes;

2. The method of claim 1, wherein obtaining a plurality of test values of a target item parameter of the interactive item during different use processes comprises:

3. The method of claim 2, wherein the data item is a target image, and the obtaining a plurality of data sets of the interactive prop generated during different use processes comprises:

4. The method of claim 3, wherein extracting a target region from each test image of the plurality of image sets, obtaining a binary image of each target region comprises:

5. The method according to claim 3, wherein before the anomaly detection is performed on each binary image and the anomaly image in each binary image is deleted to obtain each target image, the method further comprises:

6. The method of claim 3, wherein determining the plurality of test values for the target prop parameter from the respective data items of the respective data sets comprises:

7. The method of claim 6, wherein the reference point is a geometric center of a test image corresponding to the target image; or, the reference point is a pixel point of which the lowest pixel value in the target image is 1.

8. The method of claim 3, wherein the multiple tests of the interactive prop with different prop parameters via the configuration file result in multiple image sets comprising:

9. The method according to claim 1, wherein the determining the use state information and the stability information of the target item parameter according to each test value of the target item parameter comprises:

acquiring at least one of an average value or a standard deviation of each test value, and determining at least one of the average value or the standard deviation as the use state information of the target prop parameter;

and acquiring a plurality of comparison groups corresponding to a plurality of data groups in which each test value is positioned, and performing statistical analysis on the plurality of data groups and the plurality of comparison groups to obtain stability information of the target prop parameter.

10. The method of claim 9, wherein the performing a statistical analysis on the plurality of data sets and the plurality of comparison sets to obtain stability information of the parameters of the target property comprises:

performing replacement inspection on the plurality of data groups and the plurality of comparison groups to obtain a distribution result of the target prop parameter;

carrying out self-service statistics on the plurality of data groups and the plurality of comparison groups to obtain the difference degree of the target prop parameters;

determining at least one of the distribution result or the degree of difference as the stability information.

11. The method of claim 10, wherein performing a permutation test on the plurality of data sets and the plurality of comparison sets to obtain the distribution result of the target prop parameter comprises:

12. The method of claim 10, wherein the self-service statistics of the plurality of data sets and the plurality of comparison sets to obtain the degree of difference of the target property parameters comprises:

13. An abnormality detection apparatus, characterized in that the apparatus comprises:

the system comprises an acquisition module, a storage module and a display module, wherein the acquisition module is used for acquiring a plurality of test values of target prop parameters of an interactive prop in different using processes, the target prop parameters are used for reflecting target interactive characteristics of the interactive prop in the using process, the plurality of test values are determined based on a plurality of target images, and the target images are images reserved after deleting abnormal images in an image set generated in the different using processes;

14. A computer device comprising one or more processors and one or more memories having at least one program code stored therein, the at least one program code loaded and executed by the one or more processors to perform operations performed by the anomaly detection method of any one of claims 1 to 12.

15. A storage medium having stored therein at least one program code, the at least one program code being loaded into and executed by a processor to perform operations performed by the anomaly detection method of any one of claims 1 to 12.